Pending on our screen can mean the FR10 already got it

Only the plate label. It's reused when we inspect the same reference again. No reliable match for the request that was in flight.

Then how will the operator see that uncertainty after reopening? An empty result box could look like permission to start over.

We had a screen where a failed lookup looked exactly like an empty queue. People naturally treated it as nothing to do. I'd give your unknown outcome its own visible state, with a route to reconcile it rather than a retry shortcut.

Rosa, did you find the old result elsewhere, or only improve how the next failure was shown?

Draft now says outcome unknown beside the plate label. No automatic resend from that state. Developer is adding an attempt identifier and checking the handoff between saving intent and sending the request.

Sam, we recovered some results from the receiver's retained list, but not all. We left the rest unresolved. Fixing the screen didn't fill in the missing history.

Offline tests cover a crash before send, after send and before result save. The uncertain cases stay out of the send queue after reopening. Still need the end-to-end recovery check with controls.

Have you also tried a failed lookup during recovery? Rosa's empty-queue example sounds worth keeping in the tests, not just the crashes you already know about.